摘要
主要研究设备异常度的检测问题。在无故障样本的情况下,如何快速检测设备异常度已经成为状态检测的重要问题。为此,提出一种用于设备异常和异常度检测的信息匹配检测方法——可变阈值信息检测器(Variable threshold information detector,VTI-detector),在分析分散增量理论并提出数据分布相亲有限信息密度概念的基础上,计算了每个正常训练样本对自己样本(正常样本)的匹配阈值,建立了带有匹配阈值信息的矩阵,即确立VTI-detector。最后利用确立的VTI-detector,结合相亲有限信息密度的概念,提出了设备异常度的计算公式。以UCI数据库中的Iris数据集为例,将所提出的异常检测方法与其他三种常用的异常检测方法进行对比,显示VTI-detector具有比其他方法更好的检测性能。利用异常度公式在线计算轴承正常和各种故障状态的异常度,并以异常度曲线的形式进行显示,结果表明故障异常度检测效果明显,具有较好的应用前景。
The Equipment abnormal degree detection problem is studied. How to detect the abnormal degree of equipment rapidly without fault samples has become an important problem of state inspection, so a new information matching detection method called variable threshold information detector 0~TI-detector) is presented. On the basis of the disperse increment theory and proposed concept of amicable limited information density, matching threshold of each normal training sample compared to itself (normal samples) is calculated. Furthermore, vector including the information of matching threshold, namely VTI-detector, is built. At last, based on the VTI-detector and amicable limited information density concept, an abnormal degree formula is proposed. To take the Iris data sets of UCI for an example, the VTI-detector method shows a better detection performance by comparison with other three kinds of commonly used anomaly detection methods. Through the calculation of abnormal degree formula, the abnormal degree of normal and abnormal bearings can be displayed in the form of real-time curves. The result shows that this method has an obvious effect, which proves it has a good application prospect.
出处
《机械工程学报》
EI
CAS
CSCD
北大核心
2013年第8期25-31,共7页
Journal of Mechanical Engineering
基金
上海市科学技术委员会基础研究(7150080050)
高等学校博士学科点专项科研基金(20103108110006)资助项目
关键词
相亲有限信息密度
可变阈值信息检测器
异常度
Amicable limited information density Variable threshold information detector Abnormal degree